Here is why WB is not enough

The last few days we have had a friendly forum feud about white balance.
A forumer suggested that “it will be sufficient to perform a white balance, because then all colours will fall into their right places”.

I disagreed, and set out to perform a few simple experiments to see who would become the feudal winner.

First test

Question: Do the camera’s WB settings influence the end result?

Using a Colorchecker Passport as a suitable target, I shot the target in three ways:
the incoming light was identical, ISO and exposure identical, but the camera’s WB was set to

  • auto
  • daylight
  • measured WB

I whitebalanced at square #22 (the neutral grey 5 patch), and changed exposure to lift L in the Lab value of that square to as close to Lab 50.83 -0.64 -0.14 that I could manage.

Result: The coloured squares of the three shots are quite similar — but not identical.

Open this file in the Gimp, and click through the layers to see the difference between the processed shots: trio_ADM.xcf (836.5 KB) (See layer names for identification.)

Second test

Question: Does the temperature of the incoming light influence the end result?

Same settings as above, but this time, the camera’s WB was set to Auto. Incoming light = a little flash gun (TT350) set to TTL — the flash was equipped with

  • Rosco Sun filter full CTO (#85)
  • Rosco Full blue (#CTB)

I whitebalanced at square #22 (the neutral grey 5), and changed exposure to lift L in the Lab value of that square to as close to Lab 50.83 -0.64 -0.14 that I could manage.

Open this file in the Gimp, and click through the layers to see the difference between the processed shots. tada.xcf (814.2 KB) (See layer names for identification.)

The coloured squares of these two shots really differ, which indicates that proper whitebalance alone is not sufficient to “rescue” original colours.

So, after that, I believe that now we can safely declare that

  • Your camera’s WB setting is of little importance for the end result.
  • Temperature and quality of “incoming light” has a large influence.
  • Inferior light emitters – like cheap LED bulbs having gaps in their spectra – those images can never be salvaged.

Have fun!
Claes in Lund, Sweden

8 Likes

That is pretty interesting. Can you publish raw files of the second experiment with filters?

You’ll learn more if you conduct such experiments yourself. :upside_down_face:

2 Likes

The word feud was certainly tongue in cheek. We had a friendly discussion because in the following thread @HeyItsRobin was trying to work out a workflow to accurately capture colour. I was wondering if this could not be achieved by WB alone if you matched the grey patch to neutral grey. Would not the other colors set themselves correctly?

Here I have taken the three images shot by Claes and put each layer side by side so we can do a better direct comparison. To my eyes the colors look very consistent between the three shots, however, the auto WB seems to have less punch which based on the white patch may be contrast. Not sure why this would be.

Thanks Claes

2 Likes

Before making any conclusions it would be nice to see a spectrogram of the flashgun you’re using. Manufactures do not publish spectrograms for their flashlight products but I remember reading somewhere that they are not perfect as well. Unlike tungsten bulbs flashlight might also have gaps in spectrum.

1 Like

Just for reference

Full article is here
274676214_Multispectral_calibration_to_enhance_the_metrology_performance_of_C

1 Like

Would these figures suffice?
Small flash:
tt350.txt (1.0 KB)
Big flash:
V860II.txt (1.0 KB)

I used to play with gnuplot previously some time ago. Here’s what I got with your data for these two flashes. I’m not sure about accuracy of the visible spectrum though (I found rgb values somewhere on the net)

Can you give me data in the same format for daylight, halogen bulb and halogen bulb covered with Lee 201, please? I would put this data into my gnuplot script

1 Like